RF Sputtering Physics
RF sputtering enables deposition from insulating targets by using alternating electric fields at 13.56 MHz. The capacitive sheath alternates polarity, preventing charge buildup while maintaining plasma.
Self-Bias Voltage Development
V_bias = -V_RF × (A_ground / A_target)^n
n ≈ 2-4 (area ratio exponent)
Physical origin:
• Electrons mobile at RF frequency (ω > ω_pi,electron)
• Ions too heavy to follow RF (ω << ω_pi,ion)
• DC bias develops to balance electron/ion flux
Typical values:
V_RF = 100-500 V (peak-to-peak)
V_bias = -50 to -300 V (DC self-bias)
Asymmetry factor: Γ = (A_powered / A_ground)
Stronger asymmetry → larger |V_bias|
Capacitive Sheath Physics
Sheath thickness (Child-Langmuir):
s = λ_D × (2V_RF / kT_e)^(3/4)
λ_D = √(ε₀kT_e / n_e e²) (Debye length)
For T_e = 3 eV, n_e = 10¹¹ cm⁻³:
λ_D ≈ 60 μm
s ≈ 0.5-2 mm (time-averaged)
RF impedance:
Z_sheath = 1 / (jωC_sheath)
C_sheath = ε₀A / s ≈ 50-200 pF
Reactance dominates at low ω
→ Requires matching network (L-C circuit)
Electron Heating Mechanisms
Stochastic heating (collisionless):
P_stoch ∝ n_e × V_RF² × ω × (ω / ν_m)
ν_m: electron-neutral collision frequency
Ohmic heating (collisional):
P_ohmic = σ E_RF² / 2
σ = n_e e² / (m_e ν_m) (conductivity)
Dominant regime:
ω < ν_m: ohmic (high pressure)
ω > ν_m: stochastic (low pressure)
At 13.56 MHz, 5 mTorr Ar:
ν_m ≈ 10⁸ s⁻¹, ω ≈ 8.5×10⁷ rad/s
→ Transition regime
Ion Bombardment Energy Distribution
Ion energy at substrate:
E_ion = e × |V_bias| + E_plasma
E_plasma ≈ 5 × kT_e (plasma potential)
Energy spread (IEDF width):
ΔE ≈ 2 × e × V_RF × (ω / ν_i)
ν_i: ion transit frequency through sheath
Narrow IEDF: ν_i >> ω (thick sheath, slow ions)
Broad IEDF: ν_i ≈ ω (thin sheath, fast ions)
For SiO₂ deposition:
E_ion = 50-300 eV (low damage)
ΔE = 10-50 eV (moderate spread)
Impedance Matching Network
L-match configuration:
Z_load = R_plasma + jX_sheath
R_plasma ≈ V_RF² / (2P_absorbed)
Matching condition:
Z_in = 50 Ω (transmission line)
Smith chart solution:
1. Series inductor L_s → move along constant-R circle
2. Shunt capacitor C_p → move toward center
Bandwidth:
Δf / f₀ ≈ 1 / Q
Q = ω₀L / R ≈ 5-20
Δf ≈ 0.5-3 MHz (need auto-matching)
Sputter Yield for Insulators
RF sputter yield Y_RF:
Y_RF = Y_DC(E_ion,avg) × η_RF
η_RF = efficiency factor (0.5-0.8)
Reduced by:
• Time-varying sheath (ions see average potential)
• Surface charging effects
• Neutralization at insulator surface
Typical yields (500 eV Ar⁺):
SiO₂: Y ≈ 0.4-0.6 molecules/ion
Al₂O₃: Y ≈ 0.3-0.5 molecules/ion
Si₃N₄: Y ≈ 0.5-0.7 molecules/ion
ITO: Y ≈ 0.6-0.9 molecules/ion
Deposition Rate Calculation
R_dep = (Y_RF × Γ_ion × M × A_target) / (ρ × A_substrate × d²)
Γ_ion = ion flux density (ions/cm²/s)
Γ_ion ≈ n_i × v_Bohm / 4
v_Bohm = √(kT_e / M_ion)
For n_e = 5×10¹¹ cm⁻³, T_e = 3 eV:
Γ_ion ≈ 2×10¹⁶ ions/cm²/s
SiO₂ at 600 W RF:
R_dep ≈ 1-3 Å/s
(slower than metals due to lower Y)
Film Stoichiometry Control
Oxygen deficiency in oxides:
δ = (O_stoich - O_film) / O_stoich
Causes:
• Preferential O sputtering (lighter)
• Oxygen loss in flight
• Insufficient O₂ reactive gas
Solution: Reactive RF sputtering
P_total = P_Ar + P_O₂
Ratio: P_O₂ / P_Ar ≈ 0.1-0.3
Hysteresis effect:
• Metallic mode (low P_O₂): high rate, O-poor
• Oxide mode (high P_O₂): low rate, stoichiometric
• Transition mode: optimal balance
Plasma Density and Frequency
Electron plasma frequency:
ω_pe = √(n_e e² / (ε₀ m_e))
For n_e = 10¹¹ cm⁻³:
ω_pe ≈ 1.8×10¹⁰ rad/s ≈ 2.8 GHz
Condition for RF coupling:
ω_RF < ω_pe (overdense plasma)
13.56 MHz << 2.8 GHz ✓
Skin depth:
δ_skin = c / ω_pe ≈ 1-5 cm
RF fields penetrate entire plasma
Higher frequencies (27, 40 MHz):
• Higher n_e possible
• Better uniformity
• Narrower IEDF
Power Coupling Efficiency
Forward power: P_fwd (from RF generator)
Reflected power: P_ref (mismatch)
Absorbed power: P_abs = P_fwd - P_ref
Coupling efficiency:
η_couple = P_abs / P_fwd
η_couple = 1 - |Γ|²
Γ = (Z_load - Z₀) / (Z_load + Z₀)
Optimal matching:
|Γ| < 0.1 → η > 99%
P_ref < 1% of P_fwd
Poor matching:
|Γ| > 0.3 → η < 91%
→ Overheating, arc risk
Film Properties vs RF Parameters
Density: ρ_film / ρ_bulk = f(E_ion, P_Ar)
High E_ion (high V_bias):
• Atomic peening → dense films
• Compressive stress
• Smoother surface
Low E_ion:
• Columnar growth → porous films
• Tensile stress
• Rough surface
Refractive index (SiO₂):
n = 1.46 (bulk) vs 1.42-1.45 (sputtered)
Depends on density and stoichiometry
Stress control:
σ ∝ (E_ion - E_threshold) / (kT_substrate)
Optimize T_substrate and V_bias
Advanced RF Matching Techniques
Impedance matching in RF sputtering requires sophisticated network design to minimize reflected power and maximize coupling efficiency. Modern matching networks employ variable capacitors and inductors with automatic tuning capabilities.
Complex impedance:
Z = R + jX = |Z| × e^(jφ)
|Z| = √(R² + X²)
φ = arctan(X / R)
Reflection coefficient:
Γ = (Z_load - Z₀) / (Z_load + Z₀)
Z₀ = 50 Ω (characteristic impedance)
VSWR (Voltage Standing Wave Ratio):
VSWR = (1 + |Γ|) / (1 - |Γ|)
Target: VSWR < 1.2 (|Γ| < 0.1)
Acceptable: VSWR < 2.0 (|Γ| < 0.33)
Return loss:
RL = -20 × log₁₀(|Γ|) [dB]
Matching bandwidth:
BW = f₀ / Q
Q = X / R (quality factor)
For 13.56 MHz:
Q ≈ 10 → BW ≈ 1.4 MHz
Need auto-tuning for process changes
Substrate Heating and Thermal Management
RF power deposition causes significant substrate heating, affecting film properties and process stability. Understanding heat transfer mechanisms is critical for temperature control.
Heat balance equation:
P_in = P_radiation + P_conduction + P_convection
Ion bombardment heating:
P_ion = Γ_ion × E_ion × A_substrate
Radiation (Stefan-Boltzmann):
P_rad = ε × σ_SB × A × (T⁴ - T_chamber⁴)
ε: emissivity (0.1-0.9)
σ_SB = 5.67×10⁻⁸ W/(m²·K⁴)
Backside gas conduction:
P_cond = h × A × (T_wafer - T_chuck)
h = thermal conductance (W/m²/K)
h ∝ P_He (helium pressure)
Typical values:
P_ion ≈ 10-100 W
P_rad ≈ 5-50 W
P_cond ≈ 20-200 W (with He)
Steady-state temperature:
T_substrate = f(P_RF, P_He, T_chuck)
Time constant:
τ_thermal = (m × C_p) / (h × A)
τ ≈ 10-60 s (typical)
Plasma Diagnostics and Monitoring
In-situ diagnostics enable real-time process control and fault detection. Common techniques include optical emission spectroscopy (OES), Langmuir probes, and RF voltage/current monitoring.
Optical Emission Spectroscopy:
I_emission ∝ n_excited × A_ji
n_excited: excited state density
A_ji: transition probability
Corona model (low pressure):
n_excited / n_ground ≈ (k_excitation × n_e) / A_ji
k_excitation: rate coefficient
Useful transitions:
Ar I (750 nm): plasma density
Ar II (488 nm): ionization degree
Metal lines: sputter rate
O I (777 nm): reactive species
Langmuir probe:
I(V) = I_sat × [1 - exp((V - V_plasma)/(kT_e))]
Electron temperature:
T_e = e / (d ln(I) / dV)
Plasma density:
n_e = I_sat / (0.6 × e × A_probe × √(kT_e / M_i))
RF compensation:
Need filtered probes for RF plasmas
Minimize probe perturbation
Multi-Frequency RF Sputtering
Dual-frequency systems combine low-frequency (LF) and high-frequency (HF) power to independently control ion bombardment and plasma density.
Dual-frequency configuration:
f_LF = 400 kHz - 2 MHz (ion control)
f_HF = 13.56 - 40.68 MHz (density control)
Ion energy at LF:
E_ion(LF) = e × V_LF × √(ω_LF / ν_i)
Higher energy, narrow IEDF
Plasma density at HF:
n_e(HF) ∝ P_HF^α
α ≈ 0.5-0.7
Advantages:
• Independent E_ion and n_e control
• Wider process window
• Better uniformity
• Reduced charging damage
Challenges:
• Complex matching networks
• Frequency coupling effects
• Harmonic generation
• Cost increase
Applications:
• Dielectric etching
• Oxide deposition
• Charge-sensitive devices
Pulsed RF Sputtering
Pulsed power delivery enables time-averaged control of ion energy and flux, beneficial for reducing defects and improving film quality.
Pulse parameters:
f_pulse = 10 Hz - 100 kHz
Duty cycle: D = t_on / (t_on + t_off)
Time-averaged quantities:
P_avg = P_peak × D
Γ_avg = Γ_peak × D
Benefits of pulsing:
• Reduced arcing (insulators)
• Lower substrate heating
• Better stoichiometry control
• Defect reduction
Synchronous pulsing:
RF and bias pulsed together
or with phase offset
Optimal parameters:
f_pulse = 100-1000 Hz
D = 20-80%
Transient effects:
Plasma decay time: τ_plasma ≈ 1-100 μs
Ion transit time: τ_ion ≈ 1-10 μs
Require: T_pulse >> τ_plasma
for stable operation
Contamination Control and Particle Generation
Particle contamination degatively impacts yield. Understanding sources and mitigation strategies is essential for production environments.
Particle sources:
1. Target flaking (stress buildup)
2. Chamber wall deposits
3. Shield spalling
4. Arcing events
5. Pump backstreaming
Flake detachment criterion:
σ_film > σ_adhesion + σ_stress
Typical values:
σ_adhesion ≈ 10-100 MPa
σ_stress ≈ 100-1000 MPa (thick films)
Mitigation strategies:
• Periodic chamber cleaning
• Stress-optimized shields
• Target conditioning
• Pressure optimization
• Magnetic field tuning
Particle detection:
• In-situ OES monitoring
• Wafer inspection (post-process)
• Chamber pressure spikes
Cleanliness metrics:
Defect density: < 0.01 cm⁻²
Critical size: > 50 nm
Kill ratio: < 1% yield loss
Scale-Up and Manufacturing Considerations
Transitioning from R&D to production requires addressing uniformity, throughput, and reliability at scale.
Uniformity requirements:
σ/μ < 2% (thickness)
σ/μ < 3% (composition)
across 300 mm wafer
Magnetron design:
• Rotating magnet arrays
• Optimized racetrack
• Uniform erosion pattern
Target utilization:
U = Volume_eroded / Volume_total
Standard: U ≈ 20-30%
Advanced: U ≈ 40-50%
Throughput calculation:
WPH = (3600 / t_cycle) × η_uptime
t_cycle = t_load + t_pump + t_process + t_unload
Typical cycle:
t_load = 30 s
t_pump = 60 s
t_process = 120 s
t_unload = 20 s
→ WPH ≈ 15-16 (at 90% uptime)
Cost of ownership:
CoO = (Equipment + Consumables + Utilities + Labor) / WPH
Key drivers:
• Target life and cost
• Chamber cleaning frequency
• Process gas efficiency
• Yield impact
Frequency Selection and ISM Bands
The choice of 13.56 MHz for RF sputtering is driven by regulatory and physical considerations. This frequency belongs to the ISM (Industrial, Scientific, Medical) band, allowing unlicensed use for industrial processes.
ISM frequencies for plasma processing:
13.56 MHz ± 0.007 MHz (primary)
27.12 MHz ± 0.163 MHz (secondary)
40.68 MHz ± 0.020 MHz (tertiary)
Why 13.56 MHz dominates:
• Electrons respond (ω > ω_pe typically)
• Ions don't respond (ω << ω_pi)
• Commercial RF generators available
• Historical momentum
Electron plasma frequency:
ω_pe = √(n_e e² / (ε₀ m_e))
f_pe = 9 × √(n_e) Hz (n_e in cm⁻³)
For n_e = 10¹¹ cm⁻³:
f_pe ≈ 2.8 GHz >> 13.56 MHz ✓
Ion plasma frequency:
f_pi = √(M_e/M_i) × f_pe
f_pi ≈ 65 MHz << 13.56 MHz (for Ar)
Skin depth considerations:
δ = c / ω_pe
At 10¹¹ cm⁻³: δ ≈ 1-5 cm
Permits field penetration
Chamber Design and Gas Flow Patterns
Reactor geometry significantly affects process uniformity and efficiency. CFD simulations guide optimal injector placement and pumping configuration.
Gas residence time:
τ_res = V_chamber / Q_pump
V_chamber: chamber volume (L)
Q_pump: pumping speed (L/s)
For P = 5 mTorr, V = 100 L:
Q = 1000 L/s (typical)
τ_res ≈ 0.1 s
Precursor utilization:
U = (atoms deposited) / (atoms injected)
U ∝ τ_res × k_reaction × A_substrate
Typical U ≈ 5-20% (sputtering)
Most gas pumped away
Flow patterns:
• Radial injection → azimuthal uniformity
• Showerhead → axial uniformity
• Baffles → reduce channeling
Reynolds number:
Re = ρ v D / μ
Re << 1: laminar flow (typical)
Re >> 1: turbulent (rare in PVD)
Knudsen number:
Kn = λ_mfp / D
Kn > 1: molecular flow
Kn < 0.01: viscous flow
Kn ~ 0.1-1: transition
Target Cooling and Heat Management
High-power operation generates significant heat in the target, requiring active cooling to prevent melting and maintain stable operation.
Power dissipation at target:
P_target = V_target × I_discharge × η
η ≈ 0.5-0.8 (fraction absorbed)
For 5 kW at 500 V:
I = 10 A
P_target ≈ 2.5-4 kW
Cooling requirements:
Q_cool = P_target / (ρ_water × C_p × ΔT × flow)
For ΔT = 10°C:
flow ≈ 6-10 L/min (water)
Target temperature rise:
ΔT = P_target / (k × A / thickness)
k: thermal conductivity
Bonding requirements:
• Indium solder: T_melt = 156°C
• Epoxy bonding: T_max = 150-200°C
• Elastomer sealing: T_max = 200°C
Cooling channels:
• Serpentine pattern
• High flow rate (turbulent)
• Close to racetrack
Temperature monitoring:
• Thermocouples in backing plate
• IR pyrometry (if accessible)
• Interlock at T_max
Process Repeatability and Drift
Long-term stability requires understanding and mitigating sources of drift including target conditioning, chamber seasoning, and component aging.
Sources of drift:
1. Target erosion → geometry change
2. Wall coating buildup → impedance change
3. Anode shield sputtering → contamination
4. Gas delivery drift → flow variation
5. RF matchbox aging → reflection increase
Target conditioning:
Pre-sputter before production
t_condition ≈ 30-60 min
Remove oxide, contaminants
Chamber seasoning:
Deposit on walls until stable
n_seasoning ≈ 50-100 wafers
Process control strategy:
• Monitor I_discharge, V_target
• Track RF forward/reflected power
• In-situ film thickness (OES, QCM)
• Statistical process control (SPC)
Preventive maintenance:
• Target replacement: every 500-2000 wafers
• Chamber cleaning: every 1000-5000 wafers
• O-ring replacement: annually
• RF cable check: quarterly
Acceptance criteria:
• Thickness uniformity: σ/μ < 2%
• Deposition rate: ±5% of target
• Particle level: < spec limit
• Repeatability: Cpk > 1.33
Sputter Yield vs Material
Reactive Hysteresis Curve